A review of discriminant analysis in high dimensions. (9th April 2013)

• Record Type: Journal Article
• Authors: Mai, Qing
• Is Part Of: Wiley interdisciplinary reviews. Volume 5:Number 3(2013)
• Page Start: 190
• Page End: 197
• Publication Date: 2013-04-09
• Subjects: classification -- discriminant analysis -- high dimension -- ℓ1 penalty -- sparsity; Mathematical statistics -- Data processing -- Periodicals; Science -- Data processing -- Periodicals; Social sciences -- Data processing -- Periodicals; Mathematical statistics -- Periodicals; 519.50285
• DOI: 10.1002/wics.1257
• Languages: English
• ISSNs: 1939-5108

Abstract

Abstract: Linear discriminant analysis (LDA) is among the most classical classification techniques, while it continues to be a popular and important classifier in practice. However, the advancement of science and technology brings the new challenge of high‐dimensional datasets, where the dimension can be in thousands. In such datasets, LDA is inapplicable. Recently, statisticians have devoted many efforts to creating high‐dimensional LDA methods. These methods typically perform variable selection via regularization techniques. Various theoretical results, algorithms, and empirical results support the application of these methods. In this review, we provide a brief description of difficulties in extending LDA and present some successful proposals. WIREs Comput Stat 2013, 5:190–197. doi: 10.1002/wics.1257 This article is categorized under: Statistical Learning and Exploratory Methods of the Data Sciences > Clustering and Classification